Eventually consistent reads are essential reads that are performed against one node of our cluster. When an item is updated, an eventually consistent read will return either an old piece of data or a new piece of data. If our operation is not time-sensitive, then eventually consistent reads are a great approach. 

For example, if we have a game scoreboard that we record in DynamoDB and we have millions of users playing the game and viewing the scoreboard, it would be very inefficient to make sure the new scores are up to date at the exact time as they probably change quite a lot. Also, when a user sets a new high score, it is not important if the other users see this change exactly at the time it happens or a few seconds after the score was recorded.